src/scheduled_tasks.rs

   1 // Scheduler, and trait for .seconds(), .minutes(), etc.
   2 use clokwerk::{Scheduler, TimeUnits};
   3 // Import week days and WeekDay
   4 use diesel::{sql_query, PgConnection, RunQueryDsl};
   5 use lemmy_db_schema::{source::activity::Activity, DbPool};
   6 use lemmy_utils::LemmyError;
   7 use log::info;
   8 use std::{thread, time::Duration};
   9
  10 /// Schedules various cleanup tasks for lemmy in a background thread
  11 pub fn setup(pool: DbPool) -> Result<(), LemmyError> {
  12   let mut scheduler = Scheduler::new();
  13
  14   let conn = pool.get()?;
  15   active_counts(&conn);
  16
  17   // On startup, reindex the tables non-concurrently
  18   // TODO remove this for now, since it slows down startup a lot on lemmy.ml
  19   reindex_aggregates_tables(&conn, true);
  20   scheduler.every(1.hour()).run(move || {
  21     active_counts(&conn);
  22     reindex_aggregates_tables(&conn, true);
  23   });
  24
  25   let conn = pool.get()?;
  26   clear_old_activities(&conn);
  27   scheduler.every(1.weeks()).run(move || {
  28     clear_old_activities(&conn);
  29   });
  30
  31   // Manually run the scheduler in an event loop
  32   loop {
  33     scheduler.run_pending();
  34     thread::sleep(Duration::from_millis(1000));
  35   }
  36 }
  37
  38 /// Reindex the aggregates tables every one hour
  39 /// This is necessary because hot_rank is actually a mutable function:
  40 /// https://dba.stackexchange.com/questions/284052/how-to-create-an-index-based-on-a-time-based-function-in-postgres?noredirect=1#comment555727_284052
  41 fn reindex_aggregates_tables(conn: &PgConnection, concurrently: bool) {
  42   for table_name in &[
  43     "post_aggregates",
  44     "comment_aggregates",
  45     "community_aggregates",
  46   ] {
  47     reindex_table(conn, table_name, concurrently);
  48   }
  49 }
  50
  51 fn reindex_table(conn: &PgConnection, table_name: &str, concurrently: bool) {
  52   let concurrently_str = if concurrently { "concurrently" } else { "" };
  53   info!("Reindexing table {} {} ...", concurrently_str, table_name);
  54   let query = format!("reindex table {} {}", concurrently_str, table_name);
  55   sql_query(query).execute(conn).expect("reindex table");
  56   info!("Done.");
  57 }
  58
  59 /// Clear old activities (this table gets very large)
  60 fn clear_old_activities(conn: &PgConnection) {
  61   info!("Clearing old activities...");
  62   Activity::delete_olds(conn).expect("clear old activities");
  63   info!("Done.");
  64 }
  65
  66 /// Re-calculate the site and community active counts every 12 hours
  67 fn active_counts(conn: &PgConnection) {
  68   info!("Updating active site and community aggregates ...");
  69
  70   let intervals = vec![
  71     ("1 day", "day"),
  72     ("1 week", "week"),
  73     ("1 month", "month"),
  74     ("6 months", "half_year"),
  75   ];
  76
  77   for i in &intervals {
  78     let update_site_stmt = format!(
  79       "update site_aggregates set users_active_{} = (select * from site_aggregates_activity('{}'))",
  80       i.1, i.0
  81     );
  82     sql_query(update_site_stmt)
  83       .execute(conn)
  84       .expect("update site stats");
  85
  86     let update_community_stmt = format!("update community_aggregates ca set users_active_{} = mv.count_ from community_aggregates_activity('{}') mv where ca.community_id = mv.community_id_", i.1, i.0);
  87     sql_query(update_community_stmt)
  88       .execute(conn)
  89       .expect("update community stats");
  90   }
  91
  92   info!("Done.");
  93 }